jomeswang/agnes-ai-skill

This repository is an installable skill for AI coding agents (such as OpenAI Codex and similar tools) that teaches the agent how to call the Agnes AI platform's APIs for generating text, images, and video. Agnes AI is a platform that provides multiple AI models under one API key, and this skill file tells an agent which model to use for each type of task, how to authenticate, and how to structure requests. The core of the repository is a single SKILL.md file. When an agent installs this skill, it gains knowledge about three categories of Agnes models: text and chat models (agnes-2.0-flash) for conversation, code, and tool use, image models (agnes-image-2.0-flash and agnes-image-2.1-flash) for generating or editing images from text descriptions or existing images, and a video model (agnes-video-v2.0) for generating short videos, which works asynchronously and requires polling for results. The skill is intentionally kept small. It does not bundle the full Agnes API documentation or a large runtime. Instead, the execution work is split into a companion command-line tool called agnes-ai-cli, available on npm. When the CLI is installed, agents call it to make live requests rather than hand-crafting raw API calls. The skill file handles model selection and authentication guidance, the CLI handles the actual network calls, file uploads, and video polling. Installation is done through a skills installer with one command, and the repository is structured so that standard installer tools can find the skill automatically at the repository root. The README includes a gallery of example outputs: portrait photos, fashion illustrations, a perfume advertisement image, and short video clips generated with agnes-video-v2.0, including an ad film, an animated short, and a cinematic desert scene. The project is licensed under MIT. Agnes AI's free-access positioning is described as a current platform stance rather than a permanent guarantee, so checking current pricing before sustained use is advised.